ArticlePDF Available

Bioimpedance as a potential diagnostic decision tool for skin neoplasms

Authors:

Abstract and Figures

The aim is to investigate the diagnostic power of electric impedance measurements of basal cell carcinoma (BCC), benign pigmented cellular nevi (BEN), and normal skin. Impedance of 258 BEN and 34 BCC were measured from 1 to 1000 kHz. The data were analysed using receiver operating characteristic (ROC) curves. Area under the ROC curves of BEN vs. references was 0.83, BCC vs. references 0.92, and BEN vs. BCC 0.87, i.e. the impedance technique can, with some technical enhancements, be useful in classifying skin lesions.
Content may be subject to copyright.
BIOIMPEDANCE AS A POTENTIAL DIAGNOSTIC DECISION TOOL FOR
SKIN NEOPLASMS
P. Åberg*, I. Nicander**, U. Holmberg***, J. Hansson**** and S. Ollmar*
* Medical Engineering, Karolinska Institute, Stockholm, Sweden
** Dermatology, Huddinge University Hospital, Huddinge, Sweden
*** Department of Surgery, Läkarmottagningen Hötorget, Stockholm, Sweden
**** Department of Oncology-Pathology, Radiumhemmet, Karolinska Hospital and Institute,
Stockholm, Sweden
Abstract: The aim is to investigate the diagnostic
power of electric impedance measurements of basal
cell carcinoma (BCC), benign pigmented cellular
nevi (BEN), and normal skin. Impedance of 258 BEN
and 34 BCC were measured from 1 to 1000 kHz. The
data were analysed using receiver operating charac-
teristic (ROC) curves. Area under the ROC curves of
BEN vs. references was 0.83, BCC vs. references
0.92, and BEN vs. BCC 0.87, i.e. the impedance tech-
nique can, with some technical enhancements, be
useful in classifying skin lesions.
Keywords: electric bio-impedance, skin cancer, re-
ceiver operating characteristic, basal cell carcinoma,
benign pigmented cellular nevi
Introduction
Emtestam et al. [1] and Kapoor [2] used electrical
impedance to assess basal cell carcinoma (BCC), benign
pigmented cellular nevi (BEN) with normal skin as
references. It was found that there are significant differ-
ences for BCC vs. references and BEN vs. BCC. This
implies that skin impedance can be used to classify skin
lesions. The aim of this paper is to reproduce the previ-
ous results and to determine the diagnostic power of the
technique using the area under receiver operating char-
acteristic (ROC) curves.
Materials and Methods
Impedance spectra of 258 BEN lesions in 159 pa-
tients, and 34 BCC readings in 25 patients were ob-
tained using the SciBase I depth selective impedance
spectrometer [SciBase AB, Huddinge, Sweden]. There
are subdivisions of BEN and BCC that have not been
considered in this study. Reference skin was measured
ipsi-lateral to the lesions. Prior to an impedance meas-
urement, the skin/lesion was soaked with saline solution
for approximately one minute in order to increase the
contact between the site and probe and to reduce the
high impedance of the stratum corneum which otherwise
can overshadow the information from the underlying
tissue. Lesions smaller than 5 mm in diameter were not
included in this study. The lesions were subsequently
excised for histopathological diagnosis.
For two populations, e.g. references and abnormal
lesions, sensitivity and specificity can be calculated at a
certain threshold. If the threshold is moved iteratively
from minimal to maximal value, the sensitivity and
specificity will vary between 0.0 and 1.0. A plot of the
sensitivity on the x-axis and 1-specificity on the y-axis
of the iterations is called a ROC curve. The area under
the ROC curve (AUC) is an estimate of the probability
that a randomly chosen subject is correctly diagnosed,
i.e. the AUC is a representation of the diagnostic accu-
racy of the technique [3]. For an AUC of 0.5, the diag-
nostic accuracy is as low as random classification, and
AUC of 1.0 is ideal diagnostic accuracy (i.e. sensitivity
and specificity equals 1.0).
The impedance was measured in polar coordinates
(magnitude and phase) at 31 logarithmically distributed
frequencies from 1 to 1000 kHz in five depth settings,
approximately between 0.1 and 2 mm, in volumes under
the probe. (The depth penetration is tissue dependent
and will vary with thickness of the stratum corneum.
I.e., the absolute depth penetration is uncertain. How-
ever, the relative depth penetration is certain and depth
penetration increases with depth setting). In order to
study the details of the spectra, the original data were
expanded with the impedance of the complex plane (real
and imaginary part).
The data of the lesion types were analysed using
ROC curves in the SPSS software [SPSS Inc., Chicago,
Ill., USA]. AUC was calculated for each depth setting,
frequency, and impedance presentation (magnitude,
phase, real part, and imaginary part). The probability of
the AUC’s equality to 0.5 was calculated under non-
parametric assumptions.
Results
It was found that the AUCs of the depth settings
were highly correlated, and that AUCs were higher in
the low frequency region. AUCs of the real part of the
impedance were lower than the other impedance repre-
sentations, and the magnitude and imaginary parts were
highly correlated to each other. The maximal areas un-
der the ROC curves for BEN vs. references, BCC vs.
references, and BEN vs. BCC are listed in table 1. The
maximal AUCs were significantly higher than 0.5 for all
tissue types. ROC curves of the maximal separation
between the tissue types are shown in Figure 1.
Table 1. Maximal area under curve (AUCmax) ± standard
error of ROC analysis of the impedance spectra, the
probability of equality to 0.5 (P), and location of
AUCmax.
Location of AUCmax
Tissues AUCmax P Imp. Freq. Depth
BEN vs. REF 0.829± 0.018 <0.001 Imag. 1.2 kHz 2
BCC vs. REF 0.921± 0.035 <0.001 Phase 25 kHz 1
BEN vs. BCC 0.874±0.040 <0.001 Phase 1 kHz 4
Discussion
Our results are in line with previous investigations
[1,2]. In addition to the previous results, we found that
there is a systematic difference between BEN and refer-
ences. The impedance properties of the tissue types were
partially overlapping. This was most likely due to elec-
tromagnetic noise and biological variations. The Sci-
Base I is not immune to surrounding electromagnetic
noise, and it will occasionally cause fluctuations in the
measured impedance spectra. In future experiments this
will be avoided by using a new version of the imped-
ance spectrometer. The biological impedance variations
are a result of lesion differences (size, shape, subdivi-
sion, colour, thickness of stratum corneum), variations
between subjects [4], and location variations [4,5]. It is
believed that the biological variations can be reduced by
physical modifications of the probe design. Technical
developments are in progress.
Conclusions
There are clear impedance differences for BEN vs.
references, BCC vs. references, and BEN vs. BCC. This
implies that the technique, with some developments, can
be useful in clinical diagnosis of skin lesions, or at least
as a tool to more accurately identify skin lesions for
surgical excision and histopathologic evaluation. This
may lead to a reduction in unnecessary surgical biopsies.
Acknowledgements
The authors would like to thank the staff at Depart-
ment of Surgery, Läkarmottagningen Hötorget, Stock-
holm, Sweden. This work was sponsored by SciBase
AB, Huddinge, Sweden, and the Knowledge Founda-
tion, Stockholm, Sweden.
REFERENCES
[1] L. Emtestam, I. Nicander, M. Stenström and S.
Ollmar, “Electrical impedance of nodular basal cell
carcinoma: a pilot study”, Dermatology, vol 197
(4), pp. 313-316, 1998.
[2] S. Kapoor, “Bioelectric impedance techniques for
clinical detection of skin cancer”, (MSc thesis).
University of Missouri-Rolla, Rolla, MO, USA,
2001.
[3] J.A. Hanley and B.J. McNeil, “The meaning and
use of the area under a receiver operating charac-
teristic (ROC) curve”, Radiology, vol 143(1), pp.
29-36, Apr. 1982.
[4] I. Nicander, M. Nyrén, L. Emtestam and S. Ollmar,
“Baseline electrical measurements at various skin
sites, related to age and sex”, Skin Res. Technol.,
vol. 3, pp. 252-258, 1997.
[5] P. Åberg, P. Geladi, I. Nicander and S. Ollmar,
“Variation of skin properties within human fore-
arms demonstrated by non-invasive detection and
multi-way analysis”, Skin Res. Technol., 2002 (in
press)
Figure 1. ROC curves of (a) BEN vs. references, (b) BCC vs. references, and (c) BEN vs. BCC at the location of the
maximal AUC. The diagonal line marks the area where AUC equals 0.5.
... Most alternative techniques suggested are based upon optical properties of the lesions, such as dermoscopy (skin cancer diagnosis using image analysis of microscopic photos of lesions immersed in oil) [2]. Recent publications have proposed that electrical impedance can be used to assess skin cancer34567. Skin cancer identification using electrical impedance is of special interest since it is an objective, fast, straightforward, and inexpensive technique that captures fundamentally different information than optical techniques. ...
... It has been shown that electrical impedance of skin can be used to quantify and classify induced skin alterations, e.g. skin irritations [9] and allergic reactions [10], as well as skin diseases such as diabetes [11] and, as mentioned above, skin cancer34567. A review on skin impedance is given in [12]. ...
Conference Paper
Clinically relevant information in electrical skin impedance spectra is diluted by the electrical properties of stratum corneum. Eliminating stratum corneum would improve the signals from the tissue. The aim of this paper is to describe and investigate a new type of skin impedance probe with a dedicated surface structure that penetrates through stratum corneum but not into dermis, called the minimally invasive probe. Skin impedance measurements with the minimally invasive probe showed that the effect of the stratum corneum was substantially reduced compared to regular non-invasive skin impedance. The minimally invasive impedance technique is believed to facilitate skin cancer diagnosis.
... The objective of ROC analysis is to evaluate the diagnostic accuracy of a technique, described in, Zweig & Campbell 1993, Greiner et al. 2000, and used for skin impedance in e.g. Åberg et al. 2002c]. Moreover, Brown et al. used ROC analysis on impedance data to classify of cervix cancer [Brown et al. 2000]. ...
... In vivo studies of the skin have shown abnormal patterns in impedance as a result of irritation or allergic reaction [4], [5]. Clinical studies of basal cell carcinoma have also shown significant differences between the impedance of BCC and normal skin and benign lesions, but a reliable method to identify BCC using impedance is not yet available [6]- [9]. Many studies of the electrical impedance of skin have attempted to use a single measure or index, developed heuristically, to quantify differences between tissue types or conditions. ...
Article
Variations in electrical impedance over frequency might be used to distinguish basal cell carcinoma (BCC) from benign skin lesions, although the patterns that separate the two are nonobvious. Artificial neural networks (ANNs) may be good pattern classifiers for this application. A preliminary study to show the potential of neural networks to distinguish benign from malignant skin lesions using electrical impedance is presented. Electrical impedance was measured in vivo from 1 kHz to 1 MHz at five virtual depths on 18 BCC and 16 benign or premalignant lesions. A feed-forward neural network was trained using back propagation to classify these lesions. Two methods of preprocessing were used to account for the impedance of normal skin and the size of the lesion, one based on estimating the impedance of the lesion relative to adjacent normal skin and one based on estimating the impedance of the lesion independent of size or surrounding normal skin. Neural networks were able to classify measurements in a test set with 100% accuracy for the first preprocessing technique and 85% accuracy for the second. These results indicate electrical impedance may be a promising clinical diagnostic tool for basal cell carcinoma or other forms of skin cancer.
Article
Electrical bio-impedance can be used to assess skin cancers and other cutaneous lesions. The aim of this study was to distinguish skin cancer from benign nevi using multifrequency impedance spectra. Electrical impedance spectra of about 100 skin cancers and 511 benign nevi were measured. Impedance of reference skin was measured ipsi-laterally to the lesions. The impedance relation between lesion and reference skin was used to distinguish the cancers from the nevi. It was found that it is possible to separate malignant melanoma from benign nevi with 75% specificity at 100% sensitivity, and to distinguish nonmelanoma skin cancer from benign nevi with 87% specificity at 100% sensitivity. The power of skin cancer detection using electrical impedance is as good as, or better than, conventional visual screening made by general practitioners.
Article
Bio-electrical impedance spectra of skin cancer and other lesions can be assessed using both regular non-invasive probes and a novel type of microinvasive electrode system with a surface furnished with tiny spikes that penetrate stratum corneum. The aim of the study was to compare the accuracy of detection for various types of skin cancer using impedance spectra measured with these two different electrode systems in an objective way without optimising the power of discrimination. Impedance spectra of 99 benign nevi, 28 basal cell carcinomas (BCC), and 13 malignant melanomas (MM) were measured using the two electrode systems. Classification of the lesions was made using Fisher's linear discriminant, and the sensitivities and specificities of the techniques were estimated using cross-validation. The best separation between nevi and BCC was achieved using the regular non-invasive probe (96% sensitivity and 86% specificity), whereas the best separation between nevi and MM was achieved using the microinvasive electrodes (92% sensitivity and 80% specificity). Our results indicate that the choice of electrode system is application dependent.
Article
Full-text available
A representation and interpretation of the area under a receiver operating characteristic (ROC) curve obtained by the "rating" method, or by mathematical predictions based on patient characteristics, is presented. It is shown that in such a setting the area represents the probability that a randomly chosen diseased subject is (correctly) rated or ranked with greater suspicion than a randomly chosen non-diseased subject. Moreover, this probability of a correct ranking is the same quantity that is estimated by the already well-studied nonparametric Wilcoxon statistic. These two relationships are exploited to (a) provide rapid closed-form expressions for the approximate magnitude of the sampling variability, i.e., standard error that one uses to accompany the area under a smoothed ROC curve, (b) guide in determining the size of the sample required to provide a sufficiently reliable estimate of this area, and (c) determine how large sample sizes should be to ensure that one can statistically detect differences in the accuracy of diagnostic techniques.
Article
During previous studies on the electrical impedance of the skin, we formulated a set of four physical indices that could be used to distinguish between the cutaneous effects produced by different chemical irritants. We now employ the electrical impedance technique to compare the properties of different anatomical areas of the skin, using the same set of indices. Investigations were performed on 131 healthy volunteers, who were divided into four groups on the basis of age and sex. Readings of electrical impedance were taken from ten different regions over the body, and transepidermal water loss was measured for comparison. Baseline values of electrical impedance of the skin were shown to vary, depending on the site. Age was also found to exert a major influence, causing an increase in the indices related to magnitude (MIX, RIX, and IMIX) with increasing age, and a decrease in that related to phase (PIX), while sex had only a marginal effect. As with other non-invasive techniques, baseline characteristics differ from place to place over the body surface, and age is another important determining factor.
Article
Thesis (M.S.)--University of Missouri--Rolla, 2001.
Article
Previously, we have explored the use of measurements of electrical impedance and devised 4 physically distinct indices named magnitude index (MIX), phase index, real part index and imaginary part index (IMIX) from the impedance data. Our results indicated that these indices could characterize contact reactions. The goal of the present study was to use the electrical impedance method for the preoperative assessment of nodular basal cell carcinoma (BCC). We included 11 patients with a total of 12 nodular BCC, diagnosed clinically and histologically. The noninvasive measurements were performed by transepidermal water loss (TEWL) and electrical impedance. For reference, normal looking contralateral or ipsilateral skin was used. Compared to controls, the mean TEWL of BCC was increased, but this finding was not statistically significant. The electrical impedance measurements of BCC tissue revealed statistically significant changes of the impedance indices MIX and IMIX (p </=0.001). The results suggest that the measurement of electrical impedance might become useful for investigations of BCC.
Article
It is known that the properties of human skin vary locally. The purpose of this study was to investigate the properties of human volar forearms even further using advanced non-invasive techniques and numerical methods. The skin properties of human volar forearms were investigated using measurements of trans epidermal water loss and multifrequency electrical impedance. Eight sites on the forearms of 27 healthy volunteers were measured. The sites were oriented as squares, four sites on both left and right forearm, approximately 40-50 mm apart. Analysis of variance showed significant differences for epidermal water loss (P < 0.01) and the skin impedance (P < 0.001) between the inner and outer sides of the arms. Additionally, parallel factor analysis of the full skin impedance spectra also showed that there are systematic differences between right and left arm (P < 0.01). It is crucial to design skin studies carefully in order to minimise the effects of the local skin properties of human skin.